如何在删除ibdata1和ib_logfile的情况下恢复MySQL数据库

昨天，有个朋友对公司内部使用的一个MySQL实例开启binlog，但是在启动的过程中失败了（他也没提，为何会失败），在启动失败后，他删除了ibdata1和ib_logfile，后来，能正常启动了，但所有的表通过show tables能看到，但是select的过程中却报“Table doesn't exist”。

于是，建议他试试可传输表空间。

同时，自己也测试了下，确实可行。

测试版本 MySQL 5.6.32 社区版

恢复的基本步骤

1. 将原来的数据文件COPY到其它目录下。

2. 创建同名表，表结构必须保持一致。

3. 导出表空间

mysql> ALTER TABLE t DISCARD TABLESPACE;

4. 将原来的数据文件COPY回来

5. 导入表空间

mysql> ALTER TABLE t IMPORT TABLESPACE

下面的演示会略为复杂，主要是还原整个场景，并针对上述步骤中的2,4做了一个测试。

首先，创建测试数据

在这里创建两张表。之所以创建两张相同的表是为了方便后续的测试。

mysql> create table t1(id int,hiredate datetime);
Query OK, 0 rows affected (0.14 sec)

mysql> create table t2(id int,hiredate datetime);
Query OK, 0 rows affected (0.01 sec)

mysql> insert into t1 values(1,now());
Query OK, 1 row affected (0.06 sec)

mysql> insert into t1 values(2,now());
Query OK, 1 row affected (0.00 sec)

mysql> insert into t2 values(1,now());
Query OK, 1 row affected (0.00 sec)

mysql> insert into t2 values(2,now());
Query OK, 1 row affected (0.00 sec)

关闭数据库

# /usr/test/mysql-5.6.32-linux-glibc2.5-x86_64/bin/mysqladmin shutdown -uroot -p123456 -h127.0.0.1 -P3310

删除ibdata1，ib_logfile0和ib_logfile1

[root@localhost data]# cd /data/
[root@localhost data]# ls
auto.cnf  ib_logfile0  localhost.localdomain.err  mysql_upgrade_info  test
ibdata1   ib_logfile1  mysql                      performance_schema
[root@localhost data]# rm -rf ibdata1
[root@localhost data]# rm -rf ib_logfile*[root@localhost data]# ls
auto.cnf  localhost.localdomain.err  mysql  mysql_upgrade_info  performance_schema  test

重新启动数据库

# /usr/test/mysql-5.6.32-linux-glibc2.5-x86_64/bin/mysqld --defaults-file=/usr/test/mysql-5.6.32-linux-glibc2.5-x86_64/my.cnf &

并没有报错

启动过程中的日志信息如下：

# -- ::  [Warning] TIMESTAMP with implicit DEFAULT value is deprecated. Please use --explicit_defaults_for_timestamp server option (see documentation for more details).
-- ::  [Note] /usr/test/mysql-5.6.-linux-glibc2.-x86_64/bin/mysqld (mysqld 5.6.) starting as process  ...
-- ::  [Note] Plugin 'FEDERATED' is disabled.
-- ::  [Note] InnoDB: Using atomics to ref count buffer pool pages
-- ::  [Note] InnoDB: The InnoDB memory heap is disabled
-- ::  [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins
-- ::  [Note] InnoDB: Memory barrier is not used
-- ::  [Note] InnoDB: Compressed tables use zlib 1.2.
-- ::  [Note] InnoDB: Using Linux native AIO
-- ::  [Note] InnoDB: Using CPU crc32 instructions
-- ::  [Note] InnoDB: Initializing buffer pool, size = 128.0M
-- ::  [Note] InnoDB: Completed initialization of buffer pool
-- ::  [Note] InnoDB: The first specified data file ./ibdata1 did not exist: a new database to be created!
-- ::  [Note] InnoDB: Setting file ./ibdata1 size to  MB
-- ::  [Note] InnoDB: Database physically writes the file full: wait...
-- ::  [Note] InnoDB: Setting log file ./ib_logfile101 size to  MB
-- ::  [Note] InnoDB: Setting log file ./ib_logfile1 size to  MB
-- ::  [Note] InnoDB: Renaming log file ./ib_logfile101 to ./ib_logfile0
-- ::  [Warning] InnoDB: New log files created, LSN=
-- ::  [Note] InnoDB: Doublewrite buffer not found: creating new
-- ::  [Note] InnoDB: Doublewrite buffer created
-- ::  [Note] InnoDB:  rollback segment(s) are active.
-- ::  [Warning] InnoDB: Creating foreign key constraint system tables.
-- ::  [Note] InnoDB: Foreign key constraint system tables created
-- ::  [Note] InnoDB: Creating tablespace and datafile system tables.
-- ::  [Note] InnoDB: Tablespace and datafile system tables created.
-- ::  [Note] InnoDB: Waiting for purge to start
-- ::  [Note] InnoDB: 5.6. started; log sequence number
-- ::  [Note] Server hostname (bind-address): '*'; port:
-- ::  [Note] IPv6 is available.
-- ::  [Note]   - '::' resolves to '::';
-- ::  [Note] Server socket created on IP: '::'.
-- ::  [Note] Event Scheduler: Loaded  events
-- ::  [Note] /usr/test/mysql-5.6.-linux-glibc2.-x86_64/bin/mysqld: ready for connections.
Version: '5.6.32'  socket: '/data/mysql.sock'  port:   MySQL Community Server (GPL)

可见，在启动的过程中，MySQL会重建ibdata1和redo log。

登录mysql客户端，看之前创建的t1，t2是否能访问

# /usr/test/mysql-5.6.32-linux-glibc2.5-x86_64/bin/mysql -h127.0.0.1 -p123456 -uroot -P3310

mysql> use test
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Database changed
mysql> show tables;
+----------------+
| Tables_in_test |
+----------------+
| t1             |
| t2             |
+----------------+
2 rows in set (0.00 sec)

mysql> select * from t1;
ERROR 1146 (42S02): Table 'test.t1' doesn't exist

通过show tables能查看有t1表存在，但表中的具体内容则无法查看

同时，错误日志中输出以下信息

-- ::  [Warning] InnoDB: Cannot open table test/t1 from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.

将数据目录下的test目录中的t1，t2表的数据文件和表定义文件COPY到其它地方

[root@localhost test]# cd /data/test/
[root@localhost test]# ll
total
-rw-rw----  mysql mysql   Aug  : t1.frm
-rw-rw----  mysql mysql  Aug  : t1.ibd
-rw-rw----  mysql mysql   Aug  : t2.frm
-rw-rw----  mysql mysql  Aug  : t2.ibd
[root@localhost test]# mv * /backup/
[root@localhost test]# ls
[root@localhost test]# ll /backup/
total
-rw-rw----  mysql mysql   Aug  : t1.frm
-rw-rw----  mysql mysql  Aug  : t1.ibd
-rw-rw----  mysql mysql   Aug  : t2.frm
-rw-rw----  mysql mysql  Aug  : t2.ibd

登录客户端，创建t1和t2表，注意表结构和之前的必须保持一致

细心的童鞋会发现，下面的创表语句和刚开始的创表语句并不一样，列名不一致，这个其实是为了后续的测试

mysql> show tables;
Empty set (0.00 sec)

mysql> create table t1(id_1 int,hiredate_1 datetime);
ERROR 1146 (42S02): Table 'test.t1' doesn't exist

明明已经手动移除了，为什么创建表的时候还报这个错误呢？

接下来，可先执行个drop table操作

mysql> drop table t1;
ERROR 1051 (42S02): Unknown table 'test.t1'
mysql> create table t1(id_1 int,hiredate_1 datetime);
Query OK, 0 rows affected (0.07 sec)

对于t2表，我们定义一个不同的表结构，看是否可行？

mysql> drop table t2;
ERROR 1051 (42S02): Unknown table 'test.t2'
mysql> create table t2(id_1 int);
Query OK, 0 rows affected (0.01 sec)

导出表空间

mysql> ALTER TABLE t1 DISCARD TABLESPACE;
Query OK, 0 rows affected (0.00 sec)

mysql> ALTER TABLE t2 DISCARD TABLESPACE;
Query OK, 0 rows affected (0.00 sec)

这个时候，数据目录下的test目录下，数据文件没有了，只剩下了表结构文件

[root@localhost test]# ls
t1.frm  t2.frm

导入表空间

首先对t1表进行测试

在这里，测试如下两种情况

1. 新的t1.frm+旧的t1.ibd

2. 旧的t1.frm+旧的t1.ibd

第一种情况

只是将t1表的数据文件COPY回来

[root@localhost test]# cp /backup/t1.ibd .
[root@localhost test]# chown mysql.mysql t1.ibd 

导入t1表的表空间

mysql> ALTER TABLE t1 IMPORT TABLESPACE;
Query OK, 0 rows affected, 1 warning (0.21 sec)

mysql> show warnings;
+---------+------+-----------------------------------------------------------------------------------------------------------------------------------------+
| Level   | Code | Message                                                                                                                                 |
+---------+------+-----------------------------------------------------------------------------------------------------------------------------------------+
| Warning | 1810 | InnoDB: IO Read error: (2, No such file or directory) Error opening './test/t2.cfg', will attempt to import without schema verification |
+---------+------+-----------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)

查看t1表是否能访问

mysql> select * from t1;
+------+---------------------+
| id_1 | hiredate_1          |
+------+---------------------+
|    1 | 2016-08-18 17:45:02 |
|    2 | 2016-08-18 17:45:02 |
+------+---------------------+
2 rows in set (0.00 sec)

mysql> flush table t1;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from t1;
+------+---------------------+
| id_1 | hiredate_1          |
+------+---------------------+
|    1 | 2016-08-18 17:45:02 |
|    2 | 2016-08-18 17:45:02 |
+------+---------------------+
2 rows in set (0.00 sec)

喔，确实能访问，注意观察，表的列名与新的创表语句保持一致。

在这里之所以使用flush table操作，是为了刷新内存中的表定义。

下面看看t1的第二种情况，旧的t1.frm+旧的t1.ibd

mysql> ALTER TABLE t1 DISCARD TABLESPACE;
Query OK, 0 rows affected (0.00 sec)

[root@localhost test]# cp /backup/t1.frm .
cp: overwrite `./t1.frm'? y
[root@localhost test]# cp /backup/t1.ibd .
[root@localhost test]# chown mysql.mysql t1.frm
[root@localhost test]# chown mysql.mysql t1.ibd 

mysql> ALTER TABLE t1 import TABLESPACE;
Query OK, 0 rows affected, 1 warning (0.04 sec)

mysql> select * from t1;
+------+---------------------+
| id_1 | hiredate_1          |
+------+---------------------+
|    1 | 2016-08-18 17:45:02 |
|    2 | 2016-08-18 17:45:02 |
+------+---------------------+
2 rows in set (0.00 sec)

mysql> flush table t1;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from t1;
+------+---------------------+
| id   | hiredate            |
+------+---------------------+
|    1 | 2016-08-18 17:45:02 |
|    2 | 2016-08-18 17:45:02 |
+------+---------------------+
2 rows in set (0.00 sec)

第一次查询的时候还是新的列名，对表进行flush后，就恢复到原来的列名了。

下面来看看t2表的导入情况

因为t2表的表结构发生了改变，在这里，也是测试如下两种情况

1. 新的t2.frm+旧的t2.ibd

2. 旧的t2.frm+旧的t2.ibd

首先，只是导入t2表的数据文件

[root@localhost test]# cp /backup/t2.ibd .
[root@localhost test]# ll
total
-rw-rw----  mysql mysql   Aug  : t1.frm
-rw-r-----  mysql mysql  Aug  : t1.ibd
-rw-rw----  mysql mysql   Aug  : t2.frm
-rw-r-----  root  root   Aug  : t2.ibd
[root@localhost test]# chown mysql.mysql t2.ibd 

导入t2表的表空间进行测试

mysql> ALTER TABLE t2 import TABLESPACE;
Query OK, 0 rows affected, 1 warning (0.03 sec)

mysql> select * from t2;
+------+
| id_1 |
+------+
|    1 |
|    2 |
+------+
2 rows in set (0.00 sec)

mysql> flush table t2;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from t2;
+------+
| id_1 |
+------+
|    1 |
|    2 |
+------+
2 rows in set (0.00 sec)

从结果可以看出，只能读出第一列。

下面测试第二种情况，旧的t2.frm和t2.ibd

mysql> ALTER TABLE t2 DISCARD TABLESPACE;
Query OK, 0 rows affected (0.06 sec)

[root@localhost test]# rm -rf t2.frm
[root@localhost test]# cp /backup/t2.frm .
[root@localhost test]# cp /backup/t2.ibd .
[root@localhost test]# chown mysql.mysql t2.frm
[root@localhost test]# chown mysql.mysql t2.ibd 

mysql> ALTER TABLE t2 import TABLESPACE;
Query OK, 0 rows affected, 1 warning (0.09 sec)

mysql> select * from t2;
+------+
| id_1 |
+------+
|    1 |
|    2 |
+------+
2 rows in set (0.00 sec)

mysql> flush table t2;
Query OK, 0 rows affected (0.00 sec)

mysql> select * from t2;
ERROR 1146 (42S02): Table 'test.t2' doesn't exist

在重新刷新后，就出现错误了，个人感觉，这个和系统表空间中的数据字典信息有关。

实际上，后续还测试了一下，如果将hiredate的列定义为varchar，则无论是使用之前的frm文件还是之后的，在导入表空间，进行查询时，数据库直接挂掉。

mysql> create table t1(id int,hiredate varchar(10));
Query OK, 0 rows affected (0.05 sec)

mysql> ALTER TABLE t1 DISCARD TABLESPACE;
Query OK, 0 rows affected (0.00 sec)

mysql> ALTER TABLE t1 import TABLESPACE;
Query OK, 0 rows affected, 1 warning (0.03 sec)

mysql> select * from t1;
ERROR 2013 (HY000): Lost connection to MySQL server during query

结论

经过上面的一系列测试，可以看到

1. 使用可传输表空间，可以解决在删除ibdata1和ib_logfile的情况下恢复MySQL数据库，当然，本文测试的前提是数据库正常关闭下删除的ibdata1和ib_logfile。

2. 使用可传输表空间，建议新建表的表结构和原来的表结构完全一致，同时，在导入表空间前，只需COPY回原来的数据文件，即ibd。

事实上，在数据库正常关闭下删除ibdata1，会导致mysql库中的以下几张表无法访问

mysql> select table_name from information_schema.tables  where table_schema='mysql' and engine='innodb';
+----------------------+
| table_name           |
+----------------------+
| innodb_index_stats   |
| innodb_table_stats   |
| slave_master_info    |
| slave_relay_log_info |
| slave_worker_info    |
+----------------------+
5 rows in set (0.00 sec)

mysql> select * from mysql.innodb_index_stats;
ERROR 1146 (42S02): Table 'mysql.innodb_index_stats' doesn't exist
mysql> select * from mysql.innodb_table_stats;
ERROR 1146 (42S02): Table 'mysql.innodb_table_stats' doesn't exist
mysql> select * from mysql.slave_master_info;
ERROR 1146 (42S02): Table 'mysql.slave_master_info' doesn't exist
mysql> select * from mysql.slave_relay_log_info;
ERROR 1146 (42S02): Table 'mysql.slave_relay_log_info' doesn't exist
mysql> select * from mysql.slave_worker_info;
ERROR 1146 (42S02): Table 'mysql.slave_worker_info' doesn't exist

同时，错误日志中报如下信息

-- ::  [Warning] InnoDB: Cannot open table mysql/innodb_index_stats from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.
-- ::  [Warning] InnoDB: Cannot open table mysql/innodb_table_stats from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.
-- ::  [Warning] InnoDB: Cannot open table mysql/slave_master_info from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.
-- ::  [Warning] InnoDB: Cannot open table mysql/slave_relay_log_info from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.
-- ::  [Warning] InnoDB: Cannot open table mysql/slave_worker_info from the internal data dictionary of InnoDB though the .frm file for the table exists. See http://dev.mysql.com/doc/refman/5.6/en/innodb-troubleshooting.html for how you can resolve the problem.

要解决这个问题，只能重建这些表。

参考

1. http://dev.mysql.com/doc/refman/5.6/en/tablespace-copying.html

2. http://dba.stackexchange.com/questions/48166/cannot-open-table-mysql-innodb-index-stats